OHDSI Symposium 2016

Patrick Ryan

With a little help from

the OHDSI community

Congratulations!

Today is our day.

We're off to Great Places!

We're off and away!

We're out on our own. And we know what we know.

And WE are the guys and gals who'll decide where to go.

We have brains in our heads.

We have evidence to teach.

We can steer ourselves

To any destination we want to reach.

And we may not find any

epi studies we believe.

In that case, of course,

we'll head straight for an RCT.

We'll look up and down methods. Look 'em over with care.

About some we will say, “I don't choose to go there.”

With our head full of brains and our hands full of data,

we're too smart to chase down unevaluated betas.

It’s observational data

we have barely touched

With so much potential,

we can learn so much.

Here in these data, patients have voices.

They tell us what's happened,

we learn from their choices.

What happened to them?

What will happen to me?

What effects will be caused if I’m not exposure-free?

It’s these questions we have, and their answers we seek.

But sometimes what we find can seem rather bleak.

We all try our darnest to gin up responses,

But sometimes it feels like we’re spewing out nonsense.

Despite the brightest of minds and the best of intentions,

Why do we keep giving patients confusing answers to

straight-forward questions?

Shin et al, J Neurol 2014:

Different designs and different data yield different

answers to the same question

Bias and errors can happen

and frequently do

even to people as brainy

and empirical as you.

But when these things start to happen,

don't worry. Don't stew.

Calibration will guide you

to know what to do.

OH!

THE PLACES WE’LL GO!

We’ll research new methods, develop best practice.

We won’t let distractions or distractors distract us.

When we know the approach that makes the most sense,

We’ll build open-source software to make it immense.

With the community collaborating each having a say,

We’ll generate evidence the OHDSI way….

LARGE SCALE

If you’re new to the community,

Don’t worry, Don’t fret.

You won't lag behind, we'll bring you along.

To standardize data, to study disease,

to hack open-source software,

to estimate risks in a breeze.

Like collaborating on best practices, standards and rules,

and sharing in the design of open-source tools,

You’ll generate evidence that can actually be trusted

By comparing with others to see its not busted

You’ll evaluate your evidence to know its reliable

And you’ll disseminate openly to be verifiable.

These steps are a process that can be well-defined.

A reproducible science, if y’all don’t mind…

You’ll be left in the dust if you stay all alone,

Because together we'll do things you can't on your own.

We’ll be tempted to keep doing things as its been done.

Because changing behavior is never fun.

But there’s danger to keeping the status quo.

Because we now know we don’t know what thought we had known.

Publication bias and p-hacking have caused quite a pickle.

No wonder reviewers can be so darn fickle.

Schuemie OHDSI 2016:

85% of published exposure-outcome pairs have p < 0.05

There’s a different path forward that we can now take.

One outlined by Schuemie and Peter Rijnbeek.

For reliable Evidence Generation to work and succeed,

openness and transparency must be our common creed.

Sharing your protocol, posting your code,

reproducible in every step of the workflow

Discrimination is naughty, except in statistics

Where parsing signal from noise makes

a fine operating characteristic

So show me you find things that need to be found.

Show me your estimate match truth on the ground.

Use negative controls to show me your bias.

And I'll show you someone doing observational data science!

For evidence generated to be

any good,

we need evidence evaluation

to check under the hood.

And patients are waiting.

Waiting to know ‘what happened to others?’

Waiting to know ‘what will happen to me?’

Waiting to know ‘what causes what?’

Waiting for evidence to get them

out of their rut.

Patients are waiting for us!

Waiting for us to do what needs to be done

To transform our data and get on the run

To stop do studies one at a time

And start turning around evidence

right on a dime.

Somehow we must escape

All this waiting and staying.

And find a new music

we all feel like playing.

HAPPY BIRTHDAY RICH BOYCE!

A large-scale approach

Based in open science.

Together we’re all stronger

as an OHDSI alliance.

And will we succeed?

Yes! We will, indeed!

We know it now clearly

with .95 AUC.

TOGETHER, WE'LL MOVE MOUNTAINS…OF EVIDENCE!

So....

be your name Vojtech or Adler or Feeney

or Christian Reich or Michael Matheny

or Schilling or Schuemie Or Szarfman Or Shah

or Hripcsak or Rijnbeek, whoever you are

We're off to Great Places!

Today is our day!

Our mountain is waiting.

So...let’s get on our way!

NOT THE END

Join the journey

Thank you for your support!

Patient-Centered Outcomes Research Institute

(PCORI)

Eugene Washington Engagement Award

Thank you for your sponsorship!

Thank you exhibitors!

Thank you OHDSI tutorial faculty!

CDM/ETL: Rimma Belenkaya, Karthik Natarajan, Mark Velez, Erica Voss

Technology stack: Taha Abdul-Basser, Lee Evans, Karthik Natarajan, Mark Velez

Cohort definition: Juan Banda, Jon Duke, Chris Knoll, Nigam Shah

Vocabulary:

Frank DeFalco, George Hripcsak, Christian Reich

Population-level estimation:

David Madigan, Martijn Schuemie, Marc Suchard

Thank you OHDSI organizing committee!

Kristin Feeney

Jon Duke

Lisa Schilling

Christophe Lambert

Steph Reisinger

Don O’Hara

Vojtech Huser

Ana Szarfman

Gregory Fusco

Andrew Williams

Matt Levine

Thank you OHDSI collaborators!

Thank you Maura!

OHDSI Symposium 2016

Join the journey!